An increasing literature within contemporary political geography has re-emerged in recent years seeking to identify relationships between size, distance/proximity, and demographic density with electoral behavior, building on a tradition of geography that has been marginalized since the 1960s along with the important contributions of critical geography (Castro 2005; Painter and Jeffrey 2009). From the critics of André Sigfrield's ([1913] 2010) famous phrase about elections in France, "granite votes on the right and limestone on the left", the relations between vote and space were viewed with suspicion, usually classified as deterministic (Barnett and Low 2004; de Castro 2005).

Our article suggests that there are important geographic factors to be considered in electoral analysis. If temporal effect is one of the possible reasons to explain electoral behavior (Kostadinova and Power, 2007) we believe space is also necessary to understand the political phenomena (Taylor and Johnston, 1979; Racu et al. 2020). Considering geographic space as absolute, relative, and relational (Harvey 2012), we do not deny the contributions of John Agnew and many other geographers who have argued that the geographical context needs to be considered in electoral analyses. Instead, we have reevaluated spatial characteristics that have been ignored for the most part since 1970. Finally, we believe this paper does not deny or diminish the importance of research that points out the political determinants (Dalton 2008; Blais 2006; Rogowski 2014) to electoral participation. On the other hand, we do criticize different analyses that have looked for a single explanation of cause and effect. For example, Blais (2006) confirmed that the institutional variables that explain the difference in participation are overvalued. Traditional explanations sought the explanatory reason for difference in electoral participation/competitiveness in institutions. Currently, political determinants and voting behavior are more prominent, as Rogowski (2014) has already shown. We argue that in the current context of declining democratic participation in elections throughout the world (Inglehart and Welzel 2009; Dalton 2017; Souza 2019), it is necessary to expand the analysis to examine the geographic conditions of electoral behavior.

According to Fradsen (2002), it is possible to draw three theoretical explanations to understand voter behavior. First, the theory of rational voting stipulates that the individual is prepared to vote based on a calculation of gains and losses. That is, when the voter believes that his individual vote can generate effective changes, it will be worthwhile to use his/her resources (such as time) to choose a candidate and go vote. Second, the psychosocial theories (Campbell et al. 1960) of voting found that collective processes are also factors in the choice of the individual in the electoral process. Individual characteristics such as education, health conditions, and occupation are associated with a collection of identities as religion, ethnicity, race, etc. that help explain electoral behavior. Finally, institutional analyses associate participation with the electoral system (proportional, parliamentary, presidential, etc.) and point out relationships between them. Some studies, for example, associate a greater electoral participation with the proportional system of party representation than with majority systems (Franklin 1996; Blais and Dobrzynska 1998).

In general, electoral participation (turnout) and competitiveness are two primary subjects in any type of electoral analysis. The purpose of our paper, however, is also to examine them from a spatial perspective. While we measure the former by the number of people who abstain from voting, we measure the latter by the difference in votes between the first and the second candidates. To corroborate and make the data more robust, we also used the concentration of votes for the winner (shown in the main article) and the total number of political parties per election (electoral fragmentation), a well-known indicator in research on voting behavior (in Appendix B). We agree with Fowler (1993), Kjaer (2007), Lachat (2011) and Rogowski (2014) that the "elections are more competitive when parties are numerous (fragmentation)" (Lachat 2011, 648) and, because of that, "citizens in more competitive election make their voting decisions in a way which is closer to the normative ideal of political representation" (Lachat 2011, 660). On the other hand, in general, research in Political Science examines the relationship between fragmentation and participation, placing the former as an independent variable for the latter, however, this is not our case.

There are plenty of articles that analyze Costa Rican democracy under the light of voter psychosocial factors, as those that associate the system's quality with supposed ethnic homogeneity and small social and regional inequality (Ameringerm 1982); although several studies have already found this factor to be a myth (Gudmundson 1986; Booth 1998). Others studies rest on an institutional analysis that reveals unique features of the country's constitution and its political engineering (Sartori 1996), as the dismantling of the Armed Forces in the 1950s, helping to explain the development of the democratic system in the country (Booth 1998). In our decision of analyzing the relationship between territorial/electoral size and turnout/competitiveness, we are investigating spatial reasons as a factor of an individual's decision to vote.

Our approach is influenced by the empirical work of Voda et al. (2017) and Veenendaal (2013), as well as in mul-tidisciplinary literature on the importance of local scale for the functioning of the democratic system (Mabileau 1993; de Castro 2005; Chemerinsky 2008). Although electoral studies are more focused on a nation-state scale, international comparisons or even state/province units, we follow the contemporary trend of research on local scale electoral geography. Along with the change in scale, new quantitative methodologies have grown increasingly common as a tool to analyze traditional geographical characteristics (Johnston and Pattie 2006), bringing new impetus to electoral geography.

Various studies analyze whether creating new states (provinces, departments, etc.) and/or new local entities (municipalities, counties, alcaldías, cantones, etc.) would bring benefits to political functionality, in addition to the important discussions about the creation of new nation-states and regionalist movements (Agnew 1996). Some of them look to the possible relationship between municipal division and the democratic or authoritarian context of the country (Resnick 2017; Voda et al. 2017), while others examine the political consequences of those spatial processes (Swianiewicz 2010; Saarimaa and Tukiainen 2016). In the background of all these studies is their attempt to identify an ideal size for virtuous political functioning.

Exploring possible reasons for differences in political participation is fundamental to the improvement of the democratic system, especially considering that democracy is neither fixed nor permanent fixed and permanent (Gallie 1956; Bobbio 2000 [1984]; Plane and Gershtenson 2004; Dahl 2012; Rauschenbach 2014). We agree with Plane and Gershtenson (2004, 92) when they confirm the importance of "the electoral context for understanding citizen behavior" and that is why it is critical that we seek new political and geographic strategies that might improve citizen participation in decision-making processes. This is the heart of the issue in political science regarding the difference between representative democracy and participatory democracy (Pateman 1970; Pieterse 2001; Avritzer 2008), and our article argues that geographic features play a central role that cannot be neglected.

We expect a negative relationship between territorial size and competitiveness; that is, the larger the territory, the lower the turnout. We expect a similarly negative relationship for electorate size; that is, the larger the number of voters, the lower the voter turnout rate.

Optimum size for political participation is a traditional theme in philosophy and political science. Montesquieu had already stated that a larger territory² means a reduction of citizen participation, making it difficult to create a republican regime enabling despotism (Montesquieu [1748] 2010). Along the same lines, Schumpeter argued that the territorial size of the nation-state precluded the direct participation of the population (Schumpeter [1942] 2017). Benjamin Constant believed that a larger territory meant, along with other social factors as the end of slavery, the change of the notion of freedom (what he calls "freedom of the ancients and freedom of the moderns"), liberating citizens from the political function and creating the image of representation (Constant [1819] 1985). According to those philosophers, greater territory would lead the population to engage in other tasks and diminish their direct political participation.

As Downs (1957) had already stated, the costs and benefits of voting are central to understanding differences in electoral participation, especially in the comparison between social classes. We assume that political participation is the result of a series of resources, motivations and mobilization factors available to citizens. Resources include time, money, knowledge, and political skills, whereas motivations may be external (relying on electoral institutions) or internal (relying on one's own choice). Mobilization corresponds to the environment around the citizens: political parties, organizations, and community affiliation. Citizens are more willing to participate when they have more resources and when they are effectively motivated and mobilized (Verba et al. 1995; Voda et al. 2017).

We expect territorial size influences at least two of the three above. Smaller municipalities tend to decrease the resources necessary for participation, since the movement to the ballot box costs less in terms of money and time. The electoral process itself is usually conceptualized in a way that reduces the distance between citizens and polling places (Gimpel and Schucknecht 2003). Like Blais said, "it makes sense to assume that people are more prone to vote if it is easy" (Blais 2006, 116). Ballots, whether electronic or manual, are taken to the different corners of a country to diminish the resources spent in the act of voting itself. Good material conditions help what Robert Dahl (1998) calls "intrinsic equality", that is, the ability of all adults to contribute to public decisions. Without such equality of access, democracy is put at risk excluding groups and their territories from processes of representation and participation.

This is the reason why the location of participatory and representative institutions is one factor in the costs of participation, as the parliaments themselves and their ordinary meetings that could promote or diminish the existence of participants (Gimpel and Schuknecht 2003; Parkinson 2012; Azevedo 2019). The territorial size of political units can be effectively reduced through central positioning of participatory institutions and through efficient means of transportation. On the other hand, transportation helps in reducing time but it often increases monetary resources for participation. As a result, absolute territorial size is still important.

Some authors argue that, theoretically, there should be a positive correlation between citizen proximity to representative institutions and institutional accountability (Dubus et al. 2010; Jakobsen and Kjaer 2016). In other words, the greater the physical distance between representatives and the represented, the lesser the concern of public officials/institutions with the wishes of the citizens. The chances of seeing and being seen increase in a smaller territory, promoting a closer relationship between "Parliament and Square" (Bobbio, Matteucci and Pasquino 2010).

Although it is difficult to find satisfactory evidence to absolutely validate this argument, in recent years some research has emerged pointing out that micro-states tend to be more democratic than larger ones. That is, data suggests that there is a negative relationship between territorial size and quality of democracy. As Veenendaal (2013) investigated, all twenty smaller states that are part of the United Nations (except for the Kingdom of Tonga) are consolidated democracies. The participation in smaller territories is greater, although the dynamics are completely different, as we will see later.

Building on the recent work of Voda et al. (2017), we anticipated in the case of Costa Rica that the size of the electorate would directly and negatively correlate with electoral participation (i.e., participation at the polls would diminish as the number of citizens increases). Lower populations, although typically voting less ideologically (Veenendaal 2013; Keheller and Lowery 2004, 2009; Oliver, Ha and Callen 2012), tend to participate more. Factors as proximity to the representative institution and proximity to other voters are crucial to understand why participation increases with a smaller population. In addition, places with lower populations tend to be more homogeneous, stimulating voter satisfaction with local politics (Dahl and Tufte 1973; Denters 2002; Denters et al. 2014).

Voda et al.'s research on the Czech Republic illustrates a finding that counters more well-known studies, such as Fornos, Power and Garand (2004) and Pérez-Liñán (2001) about Latin American countries. Voda et al. (2017) follows Olson's ([1965] 2015) Theory of Collective Action, that states that populations with a greater number of people believe less in the power of the individual vote to change anything. In this sense, Voda et al's (2017) main argument states that individuals rationally balance the cost of participation against the possible benefits. The individual may decide that the cost of going to meetings, choosing representatives, and actively participating, is too high for the possible positive outcomes this may yield for him or her.

Olson clarifies in his Theory of Collective Action that there are differences between small groups and large groups. In the former (as, for example, a group of some companies) individuals are motivated to participate in the common interests of the group, because they know that not participating would mean their exclusion from the benefits produced by the collective. Moreover, the participation of an individual in a small group is decisive for success or failure in obtaining the desired product. This difference leads to different behaviors in the process of participation: from leaders who find motivations in individual interests (as status, economic benefits, ideological, etc.) to common people who see the costs of participation outweighing the benefits and prefer to be absent. Participation is related to the size of the group involved. In our analysis, we attempt to test the validity of Voda et al's (2017) conclusions in the Costa Rican context, and evaluate whether Fornos, Power, and Garand (2004), and Perez-Linan (2001) are still more precise about Latin American countries.

Robert Dahl and Edward Tufte (1973) are also important authors in our analysis, because they show how big and small populations influence political systems in contradictory ways. They analyzed how democracy is related to size, from the creation of U.S. federalism to the electoral process. For the authors, the demographic size of the nation-state brought new challenges to a political system devised for categorically smaller Greek city-states. The representative system and federalism were creations to lessen the constraint generated by this geographic factor.

Geys (2006) analyzed 83 studies to explain voter turnout based on aggregated data, evaluating which variables were used, the theories that justified them, the success rate of these variables, and the level of the estimated impact of each of them in explaining voter turnout. The expected effect of population size was negative for turnout because, in this line of reasoning, more voters would mean a less decisive vote from a single voter for the election's outcome.

Blais (2006) argued that less populated countries have fewer electors per elected member, which makes it easier for candidates and parties to mobilize the vote. In addition to other attributes inherently connected with demographic size, Dahl and Tufte (1973) illustrated that citizen participation and the personal sense of effectiveness are also influenced by the greater or lesser numbers of voters in a political unit. Systems defined by the authors as "very small" (those with less than 10,000 inhabitants) tend to be more homogeneous (see also Keheller and Lowery 2009; Oliver, Ha and Callen 2012), with greater incentives to respond to general ballot and, therefore, dissent less.

We expect that territorial size influences competitiveness in a positive way; that is, the larger the territory, the greater the dispersion of votes. We assert this because political representation has an obvious territoriality (de Castro 2005; Magdaleno 2010). The representatives have connections with their territories and, therefore, when a great territorial extension is at stake, it is likely that distinct interests arise and candidates from different corners of the territory compete for the position.

It is important to note that based on the work of Voda et al. (2017) we also use the number of candidates and the vote distribution as an element of calculation. As already stated above, we agree with different authors who argue that fragmentation (the amount of candidates or political parties) is a significant indicator local democratic vitality. On the other hand, we believe that using only this data is a mistake, because there can be many candidates competing and the votes remain heavily concentrated in only one of them. For this reason, we believe that the number of candidates, in and of itself, is not a satisfactory representation of political competitiveness. For this reason, the number of candidates is an important factor, but an even more important factor is the percentage difference of votes between the winner and the runner-up, as even bipartisan democracies can have competitiveness, reflective of deep democratic values.

In the aforementioned research by Veenendaal (2013), the author finds that their microstate electoral competitiveness, in addition to being smaller, presents quite differently than in the territorially large states. Richards's (1982) work on the Lighthouse Islands, Malta, and the Isle of Man, drew the same conclusions. In those cases, the added factor of a small demographic size with small territories results in increased homogeneity of political interests, thus diminishing their electoral competitiveness. In addition, it is well-known that many Pacific island democracies do not have political parties (Anckar and Anckar 2000), since "existing cleavages are less produced by ideological differences, and more as a result of 'who' about 'what' and 'as'" (Veenendaal 2013, 251). It is the geographical differentiation of interests that increases with the growth of a territory.

It is interesting to note an apparent paradox between the two hypotheses above. While, on one hand, a larger territory would reduce citizen participation due to resources and motivations, on the other, it would also increase the power of dissent and, consequently, political competitiveness. The work of Saarimaa and Tukiainen (2016) reveals the same ambiguity. It is, therefore, a greater competitiveness but with less participation. It should be remembered that those sociopolitical dilemmas are part of the very discussion of different forms of political representation, such as the pure district type, the mixed district and the proportional district, used in different ways in democracies around the planet.

Electorate size, on the other hand, can influence electoral competitiveness differently. First, as shown by Voda et al. (2017), there are typically fewer candidates per capita as the electorate increases. That is, the electoral fragmentation is reduced when the total number of those who can vote is increased, because the candidates organize themselves more ideologically in political parties. However, dissent increases, resulting in less concentration of votes for a single candidate, and a smaller difference in votes cast for the winner and the runner-up. As stated earlier, in cases where this does occur, we argue that the result is greater electoral competitiveness.

In Voda et al. (2017), the authors found that when the absolute demographic size of the political unit increases there are negative effects on the number of candidates. In analyzing post-communist municipalities in the Czech Republic, the authors concluded that the absolute and relative size of political unity is a strongly influential element in electoral behavior. For them, local elections in a more populous territory have proportionally fewer candidates than those with smaller electorates. Thus, in demographically smaller units, there are more candidates per population (Aars and Offerdal 1998; Krebs 1999; Kjaer 2007; Denters and Rose 2013; Rysavy and Bernard 2013; Saarimaa and Tukiainen 2016).

Meanwhile, Robert Dahl and Edward Tufte (1973) argue that there is a directly proportional relationship between the social cost to the individual of political dissent from prevailing opinions within the community, and population size. In demographically smaller units it is more difficult for an individual to diverge from the rest of the community, unlike the larger units where anonymity and greater heterogeneity are expected, which makes the vote more ideological and partisan, and consequently a greater dispersion of votes is expected. The greater number of candidates per capita tends to be more personalized, with higher value placed on the Executive than other representative institutions (Farrugia 1993; Sutton 2007).

In sum, the literature suggests that in political units with less population (and the Costa Rican case, fewer voters), despite a larger number of candidates per capita, there is a greater concentration of votes for a few candidates, calculated by the difference between the candidates in the first and the second places and the concentration of the votes for the winner. The examples of the micro-states again stand out: although some are multi-party (except for Palau, which does not have political parties), the voting is always concentrated in one or a few political parties (Veenendaal 2013).

Costa Rica is a small country in Central America, just 51,060 km², that in 2018, according to the Instituto Nacional de Estadística y Censos (INEC), reached its historic high mark of five million inhabitants. There are seven provinces, divided into 81 cantones (Figure 1), and 485 districts which have representative seats in their respective Town Halls (Concejos Municipales).

Figure 1 Source: Instituto Geográfico Nacional (2016). Note: In 2017, a new cantón was created in Alajuela province ("Río Cuarto"), so, now there are 82 cantones in Costa Rica.

Despite its relatively smaller demographic and territorial sizes globally and even in the context of Latin America, Costa Rica is consistently considered an exemplar democracy for the region and it was built in the second wave of democracy (O'loughlin et al. 1998). Renowned institutions of democracy measurement (see, for example, Freedom House 2018), put Costa Rica as the most solid democracy (along with Uruguay and Chile) across Latin America, and one of the most reliable on the planet. Costa Rican democracy intrigues us because it developed at the same moment when theories of Latin American politics correctly predicted several emerging dictatorships in the region.

In this sense, the country presents an interesting case study to explore socio-spatial dynamics that help explain the differences between democratic systems, especially in the politically turbulent region of Central America. Despite its well-known democratic quality, in recent years there has been a significant decrease in citizen participation in regular elections, in what many authors call the "crisis of representative democracy" (Fung and Wright 2003; Cameron et al. 2012). Analyzing Costa Rica could help us discover reasons and/or geographic solutions to explain this crisis.

For this analysis, we decided to use data from the 2016 election, since it was the first one performed only for the cantonal local scale. As the literature already points out, choosing an election where a national-level voting is also at stake, distorts the analysis of the local scale as national issues change the incentives for local voting (Milbrath and Goel 1977; Ryan 2004).

Finally, it is important to note that since the 1990s Costa Rica has undergone a process of decentralization of political power, following a worldwide trend in the late twentieth and early twenty-first century (Coppedge 2012). Thus, Costa Rica created new units for the local scale, politically transforming a country that was highly centralized until then (Rivera 1998; Ryan 2004; Alfaro 2009). From this process of decentralization, the cantonal figure gained strength. Although there is the scale of the province, the source of representation in the national legislature; this division is more administrative than political. Cantones (or municipalities) represent the second level of government of the country, followed by the districts that are considered the third and smallest political-administrative unit. Thus, Costa Rica can be understood to have a national and local scale, but not a state or provincial scale of political power (Booth 1998; Alfero-Redondo 2008).

According to Ryan (2004), the virtuous economic and political effects associated with decentralization (which he pointedly criticizes), directly influenced Latin America, whose countries sought in decentralization a solution to the crisis of the decadas perdidas ("lost decades") of the 1980s and 1990s. In this regard, the cantonal election of 2016 also offers a way to examine if there has been any knowledge of the Costa Rican citizen with that political-electoral change of the country.

Our analysis uses data from multiple sources. These sources include the details of 82 cantones of the country's seven different provinces: San Jose, Alajuela, Cartago, Heredia, Guanacaste, Puntarenas, and Limón. In total, there are twelve sets of data from each canton (totaling 984 data sets), that were acquired in the 2016 Atlas de Desarrollo Cantonal de Costa Rica and the electoral data from the Tribunal Supremo de Elecciones (TSE) website. It is important to note that the universal participation rate in the 2016 elections was 35 % (TSE 2016), but its geographical distribution is very different, as we see in our analysis.

For each canton, we gather information of (1) abstentionism, (2) the number of valid votes, (3) the number of null votes, (4) the number of votes for the winning candidate, (5) the number of votes for the runner-up, (6) the number of citizens entitled to vote, (6) the number of competing parties or coalitions, (8) territorial size, (9) population size, (10) schooling rate, (11) human development index (HDI), and (12) demographic density. New proportional data using this information were added to the database.

Our sample is defined by a cross-section based on the Costa Rican 2016 elections, where the unit of analysis that determines each observation is the cantón. Thus, we have 82 observations for each of the dependent and independent variables. We reiterate that we seek possible relations between size (both territorial and electorate) and voter turnout and electoral competitiveness. Voter turnout is measured by the abstentionism rate, and electoral competitiveness is measured by the most often used proxy indicator, the difference in votes between first and second place candidates (disputed contests are settled by dividing by the total number of votes in the territory) (Geys 2006; Souza 2019). Data about concentration of votes for the winner and fragmentation are used as indicators of democratic robustness check. We have chosen these data because Regidor is the Executive Power of the canton, that is why we cannot replicate the data and methodology used by other authors (Rae 1967; de Carvalho 2003; Magdaleno 2010; Lachat 2011).³

To remove the effects of local characteristics of each province from the betas of the other independent variables, we added dummy variables for each one of the analyzed provinces. Taken together to avoid perfect multicollinearity, the dummy variable for the province of San José was not placed in the regressions and it was thus used as the reference category in the analyses of these dummies (Suits 1957). The methodology of using dummies in Ordinary Least Squares (OLS) regressions requires that we take one of the provinces out of the regression to be the element of comparison for the remaining ones. Thus, we selected San José due to its political and economic centrality.

The used control variables included the HDI, schooling rate, and population density at the cantón level. Although not the object of study, these variables (used here as a control) have already demonstrated significant influence in the main issues analyzed in this article (participation and electoral competitiveness). Therefore, we include them in the equation so that their effects do not affect the betas' estimates for the main variables (Table 1). These control variables were selected according to the literature on electoral behavior, as noted in Lappie and Marschall (2018).

Table 1

Although we work with a cross-section that includes data from just one election, the diversity that the 82 cantones present is noteworthy for their ability to produce estimators in their regressions. For example, the number of voters among the 82 cantones varies from 4,365 to 231,682 people. In addition, it is important to note that the variables "abstentionism", "number of candidates", "winner votes", "number of eligible voters", and "rate of schooling" entered the database with numbers between 0 and 100, representing percentages.